DOE's NERSC Center Deploys 10 Teraflop/s IBM Supercomputer

BERKELEY, Calif. — The National Energy Research Scientific Computing (NERSC) Center, funded by the U.S. Department of Energy's Office of Science, put its 10 teraflop/s (10 trillion calculations per second) IBM POWER-based supercomputer into service last week, providing researchers across the country with the most powerful computer for unclassified research in the United States.

The IBM supercomputer, which comprises 6,656 processors, entered production a month ahead of schedule, meaning that the system will provide up to 4 million more processor hours of computing time in the current fiscal year. The NERSC Center, located at Lawrence Berkeley National Laboratory, serves more than 2,000 researchers at national laboratories and universities across the country.

In November, the NERSC Center announced that its next supercomputer to be deployed would be a doubling of its existing IBM SP, which then had 3,328 processors. By taking this cost-effective approach, the NERSC Center was able to more quickly meet the demands of its national user community and do it with a computer architecture they were already using.

"Our partnership with IBM continues to pay significant dividends for our users, who are taking advantage of the increase in computing power to analyze and simulate scientific problems of greater complexity and size," said NERSC Center Division Director Horst Simon. "With increasing demand for computing time from research groups in DOE's SciDAC Program (Scientific Discovery through Advanced Computing), being able to provide a much larger yet robust system is critical. This is a very cost-effective way to accelerate scientific discovery throughout DOE Office of Science research programs."

Before putting the expanded IBM supercomputer into production on March 3, NERSC thoroughly tested it by running both scientific codes and benchmarking applications. The system was available more than 98 percent of the time during testing and a NERSC Center team ran benchmarks that ran at 72 percent of peak speed, much higher than that achieved on similar parallel systems.

"IBM designs systems that are equally at home running corporate databases as they are solving complex scientific problems," said Surjit Chana, IBM eServer pSeries vice president. "With the IBM supercomputing power, NERSC will have the scalability, speed and performance to continue making strides in solving the most challenging problems." The new system will include 7.8 terabytes of aggregate memory and a Global Parallel File System with 44 terabytes of storage. The system will be supported by NERSC's High Performance Storage System (HPSS), which provides 8.8 petabytes of archival data storage capacity.